Current predictions from AWI-CMIP6 (6 variables)

1 Context

Considering the review from Science Advances received in February 20222, I explored the Global Climate Models from IPCC CMIP6, despite coarser spatial resolution that could not match the coastal behaviour of Arctic whales.

Three GCM at a 25 km resolution were found (“AWI-CM-1-1-MR”,“HadGEM3-GC31-MM”,“CNRM-CM6”), including Sea Surface Temperature (tos), Sea Surface Salinity (sos), Sea Surface Height (zos) and Mixed Layer Depth (mlotst). These 4 dynamic variables were first (i) extracted from CMIP6 for the historical period (1993-2014) and contemporary period (2015-2020) from ssp126. Then, each variable was included in the initial models run on the available dataset (current predictions: “present”).

The results here are presented for summer only.

2 Model performance

Three new dynamic variables were added to each model (SSS, MLD and SSH), which improved the performance of the algorithms on the tracking data (“present distribution”).

2.1 Bowheads

Figure 2.1: Boxplot of the performance metrics for bowheads in summer (AWI-CM-1-1-MR).

2.2 Narwhals

Figure 2.2: Boxplot of the performance metrics for narwhals in summer (AWI-CM-1-1-MR).

2.3 Belugas

Figure 2.3: Boxplot of the performance metrics for belugas in summer (model: AWI-CM-1-1-MR).

The Random Forest suffers from overfitting, so the best algorithm for each species is: GBM for belugas, narwhals East and bowheads West, BRT for narwhals West and MARS for bowheads East.

3 Variable importance

Figure 3.1: Boxplot of the performance metrics for belugas in summer.

Most important variables are:

• distance to coast for belugas and narwhals West;

• SSH for narwhals East;

• SSS for bowheads West;

• SST for bowheads East.

4 Predictions Belugas

4.1 BRT

Figure 4.1: mean prediction maps for Belugas in summer (present) from the BRT.

4.2 MARS

Figure 4.2: Mean prediction maps for belugas in summer from the MARS.

4.3 GBM

Figure 4.3: Mean prediction maps for belugas in summer from the GBM.

4.4 Max GBM

Figure 4.4: Max prediction maps for belugas in summer from the GBM.

4.5 GAM

Figure 4.5: Mean prediction maps for belugas in summer from the GAM.

4.6 GLM

Figure 4.6: Mean prediction maps for belugas in summer from the GLM.

4.7 RF

Figure 4.7: Mean prediction maps for belugas in summer from the RF.

4.8 Max RF

Figure 4.8: Max prediction maps for belugas in summer from the RF.

5 Predictions Narwhals

5.1 BRT

Figure 5.1: Mean prediction maps for Narwhals in summer (present) from the BRT.

5.2 MARS

Figure 5.2: Mean prediction maps for Narwhals in summer from the MARS.

5.3 GBM

Figure 5.3: Mean prediction maps for Narwhals in summer from the GBM.

5.4 GAM

Figure 5.4: Mean prediction maps for Narwhals in summer from the GAM.

5.5 GLM

Figure 5.5: Mean prediction maps for Narwhals in summer from the GLM.

5.6 RF

Figure 5.6: Mean prediction maps for Narwhals in summer from the RF.

5.7 Max GBM

Figure 5.7: Max prediction maps for Narwhals in summer from the BRt West and GBM East.

6 Predictions Bowheads

6.1 BRT

Figure 6.1: Mean prediction maps for Bowheads in summer (present) from the BRT.

6.2 MARS

Figure 6.2: Mean prediction maps for Bowheads in summer from the MARS.

6.3 Max MARS

Figure 6.3: Max prediction maps for Bowheads in summer from the MARS.

6.4 GBM

Figure 6.4: Mean prediction maps for Bowheads in summer from the GBM.

6.5 Max GBM

Figure 6.5: Max prediction maps for Bowheads in summer (present) from the GBM.

6.6 GAM

Figure 6.6: Mean prediction maps for Bowheads in summer from the GAM.

6.7 GLM

Figure 6.7: Mean prediction maps for Bowheads in summer from the GLM.

6.8 RF

Figure 6.8: Mean prediction maps for Bowheads in summer from the RF.